Neural network-assisted denoising in attosecond transient absorption spectroscopy

Area of Science:

• Ultrafast Science
• Quantum Optics
• Spectroscopy

Background:

• Transient absorption spectroscopy is crucial for studying ultrafast dynamics.
• Conventional methods require lengthy data acquisition, limiting experimental throughput.
• Noise in measurements hinders sensitivity and accuracy.

Purpose of the Study:

• To develop a novel denoising approach for attosecond transient absorption spectroscopy.
• To improve signal sensitivity and reduce acquisition times.
• To leverage machine learning for enhanced spectroscopic measurements.

Main Methods:

• Utilized a predictive neural network for denoising spectroscopic data.
• Exploited correlations between near-infrared and high-harmonic radiation for spectral prediction.
• Simultaneously acquired pump-on and predicted reference spectra.

Main Results:

• Achieved significant noise reduction, approaching the detector's shot-noise limit.
• Enhanced signal sensitivity by an order of magnitude compared to conventional methods.
• Reduced typical multi-hour acquisition times substantially.

Conclusions:

• The predictive neural network approach offers a powerful tool for denoising attosecond transient absorption data.
• This method drastically improves measurement speed and sensitivity.
• Enables faster and more efficient investigations of ultrafast phenomena.